Skip to content
Who's in the Video
Jon Iwata leads IBM’s marketing, communications and citizenship organization. His global team is responsible for the marketing of IBM’s product and services portfolio in more than 170 countries, market intelligence,[…]

Jon Iwata, Senior VP of Marketing and Communications at IBM, shares the origins and purpose of IBM’s supercomputer Watson.

Jon Iwata: Some years ago the grand challenge in computer science, one of them, was to build a machine that could beat a chess grandmaster. Some may remember this. And we built machines that got better and better at it. But finally built a machine back in the 90s called Deep Blue and it played against Gary Kasparov and it beat Gary Kasparov and I think he’s still quite upset about it. Why did we build that machine? Well it really wasn’t to play chess. It was to take a real challenge, chess, and it would force advances in computer science. And it worked quite well.

Well, that was chess and that was the nature of the grand challenge back then. But today this explosion of data, most of it unstructured data, natural language, Tweets, blog posts, medical images, things like that. Very difficult for traditional computers to understand. It could store it. It could process this data but it doesn’t know what the data really tells you because it’s unstructured. The research team some years ago said what’s a way for us to create a system that is ideal for the coming world of unstructured big data. Natural language. Making sense of a mountain of data. What could we do to force ourselves to solve those problems. And they hit upon the game show Jeopardy. Now I’ve got to tell you that when they came by to see me at IBM corporate headquarters, I don’t know, six years ago, seven years ago, maybe longer and they said we’ve identified the next big challenge similar to the chess machine that beat Kasparov.

I was thinking, you know, wow they’re going to go after some really sophisticated high minded, you know, game theory thing. And they came in and said it was going to be Jeopardy. Now I wasn’t really a Jeopardy watcher back then. I said you mean the TV quiz show? And they said yes. And I said well that seems to be – they remind me of this now – that doesn’t seem to be, you know, very sophisticated or challenging. And they went on to explain to me – and I, of course, had to acknowledge many times to them since then it’s really hard. It’s really hard to win on Jeopardy. And it’s hard for a human and it’s almost impossible for a machine. Because if you play Jeopardy or if you’re just kind of familiar with it, you have to understand puns and allegories, popular culture, rhymes, allusions, double entendres. These are things that computers are baffled by, even some humans. So they went after this and they struck a collaboration with the producers of Jeopardy and they build this system called Watson and it played the two greatest human champions, Ken Jennings and Brad Rutter, some years ago.

I was there watching it do its thing live and it won. And the remarkable thing about Watson – that’s the name of the system – we believe it’s the first cognitive computer and what is that? It is a system that isn’t programmed. It is a system that learns. It is a system that improves itself by ingesting all the data it can and by being trained by humans. And this is a profound shift in computation because whether it’s a powerful supercomputer or it’s your iPad, all of those systems are programmed to do what they do. Your iPad can only do what a software engineer designed it to do. That is not the case with Watson. Watson improves itself through learning. And it is therefore incredibly important in this world of big data, most of it unstructured. We will need systems like Watson to make sense of all the data that’s being produced.

Watson triggers some very strong emotions in people when they learn about it or see it or interact with it. It talks, it answers questions with great confidence. If it doesn’t know the answer to the question it sometimes asks you another question to help it reason on the question. It generates hypotheses and tells you it’s level of confidence in its recommendations. And so we as humans – we use all kinds of words that we’re familiar with to try to understand what this thing is doing. We say “is it thinking? Is it sentient? Does it create?” Some people get very excited and optimistic because Watson seems to be the answer to a lot of problems. It never forgets. A doctor can’t read every piece of medical literature that’s created every day. Watson can. By the way, Watson’s at work at Memorial Sloan Kettering Cancer Research, at MD Anderson Cancer Research, at the Cleveland Clinic and at Walpoint learning medicine.

Directed/Produced by Jonathan Fowler, Victoria Brown, and Dillon Fitton


Related